Activity
From 14.01.2016 to 12.02.2016
12.02.2016
- 14:49 Task #3693 (Closed): Influence of the number of normal examples used for training on detection accuracy
- Experiments carried out also for voting detector.
11.02.2016
- 17:27 Task #3684 (Resolved): Redmine issues
- 17:23 Task #3716 (Closed): Enable to view speech signal and spectrogram in WebProkus
- 15:39 Task #3716 (Resolved): Enable to view speech signal and spectrogram in WebProkus
- done.
If you see it wrong in browser, press Ctrl-R to refresh cache. - 17:22 Task #3728 (Closed): Enable to view and edit phonetic transcription
- 13:34 Task #3728 (Resolved): Enable to view and edit phonetic transcription
- Done.
08.02.2016
- 11:00 Task #3719 (Closed): Listening test on voiced/unvoiced "ř"
- 09:23 Task #3719 (Resolved): Listening test on voiced/unvoiced "ř"
- We have gone through all the items in the listening test and commented on some of them in the test environment. More ...
- 10:58 Task #3734 (Closed): Evaluate analysis of unexpected results of listening test on voiced/unvoiced "ř"
- Daniel, try to evaluate Radek's analysis (#3720) with respect to our ultimate goal - whether to separate voiced/unvoi...
- 10:53 Task #3720 (Closed): Analyze unexpected results of listening test on voiced/unvoiced "ř"
- 10:41 Task #3720 (Resolved): Analyze unexpected results of listening test on voiced/unvoiced "ř"
- In general, it is important to realize that the voicing in /ř/ is the least stable of all phonologically voiced obstr...
04.02.2016
- 10:54 Task #3728 (Closed): Enable to view and edit phonetic transcription
- It would be nice to be able to view and edit phonetic transcription of the input text.
The procedure would be the ...
02.02.2016
- 17:14 Task #3720 (Assigned): Analyze unexpected results of listening test on voiced/unvoiced "ř"
- 17:14 Task #3719 (Assigned): Listening test on voiced/unvoiced "ř"
- 09:47 Task #3717 (Closed): Enable to view and use candidates of selected units
- 09:45 Task #3716 (Assigned): Enable to view speech signal and spectrogram in WebProkus
- The spectrogram view is still quite small. Please try to enlarge the spectrogram (and reduce the waveform view).
- 09:41 Task #3715 (Closed): Create basic version of WebProkus
01.02.2016
- 18:10 Task #3715 (Resolved): Create basic version of WebProkus
- Hotovo, na wiki jsem dal odkaz na registraci
- 18:09 Task #3717 (Resolved): Enable to view and use candidates of selected units
- 18:09 Task #3716 (Resolved): Enable to view speech signal and spectrogram in WebProkus
29.01.2016
- 14:56 Task #3709: Merge ASF files (segmentations) and SNT files (annotations)
- Two more voices were processed: KI and AJ.
*NOTE:* Several specific corrections had to be done manually. Thus, the n...
28.01.2016
27.01.2016
- 12:25 Task #3720 (Closed): Analyze unexpected results of listening test on voiced/unvoiced "ř"
- Results of so-far conducted listening tests show some inconsistencies in what was expected versus what the listeners ...
- 12:05 Task #3719 (Closed): Listening test on voiced/unvoiced "ř"
- Please carry out listening test on voiced/unvoiced "ř":
http://artic-tests.zcu.cz/autotesty_noUnitSubsts
You wi...
26.01.2016
- 15:25 Task #3713 (Resolved): Send link to listening test on voiced "ř" [R] vs unvoiced "ř" [Q] to phoneticians
- 15:24 Task #3713: Send link to listening test on voiced "ř" [R] vs unvoiced "ř" [Q] to phoneticians
- Listening tests web address is: http://artic-tests.zcu.cz/autotesty_noUnitSubsts/
The following systems were compa...
25.01.2016
- 17:25 Task #3693: Influence of the number of normal examples used for training on detection accuracy
- Starting experiments with the number of normal examples used for training of single estimators
- 17:23 Task #3693: Influence of the number of normal examples used for training on detection accuracy
- Experiments with the number of normal examples used in cross validation.
- 16:25 Task #3709: Merge ASF files (segmentations) and SNT files (annotations)
- Scripts for this task are placed in SVN repository @ARTIC_UTILS/trunk/hmm_synth/LabLight@.
A simple description of...
22.01.2016
- 15:40 Task #3717 (Closed): Enable to view and use candidates of selected units
- 15:35 Task #3716 (Closed): Enable to view speech signal and spectrogram in WebProkus
- Enable to view speech signal and spectrogram in WebProkus
- 15:32 Task #3715 (Closed): Create basic version of WebProkus
- Create basic version of WebProkus that enables:
* to synthesize any text
* to save the resulting synthetic speech... - 15:27 Task #3714 (Closed): WebProkus - software for analysis of synthetic speech
- Parent task for the design and development of "WebProkus" - web based version of Prokus
- 15:18 Task #3713 (Closed): Send link to listening test on voiced "ř" [R] vs unvoiced "ř" [Q] to phoneticians
- 11:00 Kick-off project meeting
- Kick-off project meeting in the NTIS centre.
21.01.2016
- 17:29 Task #3693 (Assigned): Influence of the number of normal examples used for training on detection accuracy
- Script preparation
- 17:28 Task #3692 (Closed): Influence of the number of anomaly examples used for crossvalidation on detection accuracy
- Experiment documented
- 17:27 Task #3687 (Closed): Propose single anomaly detectors
- 17:26 Task #3691 (Closed): Propose combination of several detectors - voting detector
- 16:30 Task #3709: Merge ASF files (segmentations) and SNT files (annotations)
- Use upper-case chars in words in ASF (copy from SNT).
- 12:23 Task #3709: Merge ASF files (segmentations) and SNT files (annotations)
- Add missing non-speech events from SNT into ASF with zero duration.
- 11:06 Task #3709 (Closed): Merge ASF files (segmentations) and SNT files (annotations)
- Merge ASF files (segmentations) and SNT files (annotations):
# Find differences between words in ASF and SNT and upd...
19.01.2016
- 21:32 Task #3684: Redmine issues
- Parent task for Redmine-related issues
- 21:11 Task #3676: RA3a - Context definition and penalisation matrix
- Parent task for context definition and penalisation matrix
*Estimated time schedule:* 01/2017 - 12/2017 - 20:56 Task #3704 (New): Detection and correction of prosodic structures
- Parent task for both detection and correction of errors in prosodic structure marking
- 14:24 Task #3699: Compute features (MFCC, LPC, LPCenv and FFTpow)
- Features computed (not all yet, though, just those following the paper, i.e. 20-20 (20msec windows, 20msec shift) MFC...
15.01.2016
- 18:26 Task #3692: Influence of the number of anomaly examples used for crossvalidation on detection accuracy
- Code for evaluation done.
- 16:40 Task #3692: Influence of the number of anomaly examples used for crossvalidation on detection accuracy
- Experiments done.
- 16:36 Task #3699 (Closed): Compute features (MFCC, LPC, LPCenv and FFTpow)
- For the whole DB compute features like:
* pitch-asynchronous/pitch-synchronous MFCC
* pitch-asynchronous/pitch-syn... - 16:31 Task #3698 (Closed): Experiment with one-class clasification for join cost enhancements
- Based on the paper attachment:ssgpub_occ_ieeesigproclet_2010.pdf, design experiment to verify the reported results.
... - 12:34 Task #3697 (Assigned): Electromagnetic articulography (EMA) based research
- "Parent" issue for EMA-based research
14.01.2016
- 17:15 Task #3692 (Assigned): Influence of the number of anomaly examples used for crossvalidation on detection accuracy
- Preparing data and scripts for the experiment
- 10:13 Task #3692 (Closed): Influence of the number of anomaly examples used for crossvalidation on detection accuracy
- Find out influence of the number of anomaly examples used for crossvalidation on detection accuracy
- 10:15 Task #3693 (Closed): Influence of the number of normal examples used for training on detection accuracy
- Find out influence of the number of normal examples used for training on detection accuracy
- 10:10 Task #3691 (Closed): Propose combination of several detectors - voting detector
- Propose combination of several detectors - voting detector
* single detectors "vote" for each example to be anomaous... - 10:03 Task #3687 (Assigned): Propose single anomaly detectors
- 10:02 Task #3690 (Closed): Annotation error detection
- Research related to the detection of annotation errors (mostly related to gross segmentation errors) in TTS corpora.
Also available in: Atom